Purpose: The relationship between circulating selenium and diabetes mellitus (DM) remains inconsistent. Therefore, the relationship between circulating selenium and DM was investigated in the present study. Patients and Methods: All participants (aged ≥18 years) were included from the National Health and Nutrition Examination Survey (NHANES) 1999-2006. Selenium concentrations from the fasting serum samples were determined using inductively coupled mass spectrometry, then grouped into quartiles. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by using multivariate logistic regression analysis and the results were stratified by age and sex. Results: A total of 2,903 (61.9±13.7 years old) participants (49.3% males) were enrolled, and 580 (19.97%) of them had DM. The mean levels of selenium were 136.4±19.6 µg/L. Patients with DM (138.76±20.02 vs 135.88±19.44, P=0.002) had higher selenium levels compared to those without DM. The OR for DM was 1.12 (95% CI=1.01-1.24; P=0.0270) for each 10 µg/L increment in selenium, and subjects in the highest quartile of selenium levels (>147.00 uµg/L) had 2.82 (95% CI=1.55-5.11; P=0.0007) times higher risk of DM compared to the lowest quartile of selenium levels. Subgroup analysis showed that selenium was independently associated with DM only in female aged <65 years. Conclusion: Circulating selenium levels were positively associated with the odds of DM, but difference in sex and age.